The use of "soft bodies" to test evolutionary simulation turns out to be surprisingly amusing.
While it may be the beginning of another long week, check out these soft robots flopping around -- it will surely brighten things up a bit.
This is the work of Nick Cheney, Robert MacCurdy, Jeff Clune, and Hod Lipson, who are members of the Cornell Creative Machines Lab and the Evolving Artificial Intelligence Lab at the University of Wyoming. They didn't set out to make a wonderful video for entertainment purposes; they were researching how soft body dynamics can improve software evolution simulations.
They started out by observing that although evolutionary algorithm simulations have been around for many years and our computational power has grown tremendously, we're still basically seeing the same stuff over and over. This, they hypothesize, is mainly due to the rigid dynamics employed in the simulations.
They set out to experiment with newer methods that allow for multiple materials to be used in the simulation, including "soft bodies." Their simulation gives the computer four types of tissue to choose from: two muscles that expand and contract at opposite intervals, a bone material, and a soft tissue material that only serves to connect other things in a flexible manner. The simulation places these materials, or "voxels," using Compositional Pattern-Producing Network, or CPPN. This gives them homogenous chunks of material as opposed to just random blocks.
Three examples of the evolving soft bodied creatures.
The results, displayed in the video using softbody simulation software called VoxCAd are not only amusing, but actually a good representation of how much more diverse the evolved creatures when given soft body pieces to use. It's at this point that you probably should forget about getting any real work done and clear your schedule for the day to play with VoxCad, a free download.
If you are interested in learning more about this research, and how the developers actually measured success and failure in the evolving "Creatures," download the PDF of their research for further reading.